Darcs versus git slides
Last night was Glasgow.pm's second technical meeting, and I finally gave a version of the DVCS-comparison talk I've been thinking about doing for at least a year. I could have done with a lot more rehearsal, something that would have been helped by finishing off the slides before (checks git-log) 5.33pm - the meeting started at 6.15 - but I think it went OK.
The idea of the talk was to
By the way, there's a useful consequence of git's design which neither mjd's slides nor mine mention, but which I mentioned in the talk, namely the reflog. It's a list of every commit you've visited, and when you were there. This means you can say things like "Show me the state of the repository at 11.15 last Monday"; more usefully, it lets you track down and recover commits that have been orphaned by some botched attempt at history rewriting. This is not a feature that you need often, but when you do need it it's an absolute lifesaver. Git's "directed graph of snapshots" model makes this feature almost trivial to add (and because git's built on a content-addressable filesystem, jumping to those orphaned commits is fast), but darcs' "bag of patches" model makes it much harder to add such a feature (though they're thinking about possible approaches that make more sense than storing your darcs repo in git).
Thanks very much to Eric Kow and Guillaume Hoffman for answering my questions about darcs. Any errors remaining are of course my own.
Anyway, you can get the slides here (slightly cleaned-up). Please let me know if they don't make sense on their own. Or, if you really care, you can browse the history here. To build the TeX source, you'll need the prosper, xypic and graphicsx packages installed.
Edit: some of the people on darcs-users were kind enough to comment, prompting some further thought. I've written a followup post in which I respond to some of the things they said.
The idea of the talk was to
- explain a little bit of what git's doing under the hood
- explain how the darcs model differs
- cut through some of the tangle of conflicting terminology
- explain why git users think branching's such a big deal, and why darcs users think that cherry-picking's such a big deal (spoilers: the answers are the same!).
By the way, there's a useful consequence of git's design which neither mjd's slides nor mine mention, but which I mentioned in the talk, namely the reflog. It's a list of every commit you've visited, and when you were there. This means you can say things like "Show me the state of the repository at 11.15 last Monday"; more usefully, it lets you track down and recover commits that have been orphaned by some botched attempt at history rewriting. This is not a feature that you need often, but when you do need it it's an absolute lifesaver. Git's "directed graph of snapshots" model makes this feature almost trivial to add (and because git's built on a content-addressable filesystem, jumping to those orphaned commits is fast), but darcs' "bag of patches" model makes it much harder to add such a feature (though they're thinking about possible approaches that make more sense than storing your darcs repo in git).
Thanks very much to Eric Kow and Guillaume Hoffman for answering my questions about darcs. Any errors remaining are of course my own.
Anyway, you can get the slides here (slightly cleaned-up). Please let me know if they don't make sense on their own. Or, if you really care, you can browse the history here. To build the TeX source, you'll need the prosper, xypic and graphicsx packages installed.
Edit: some of the people on darcs-users were kind enough to comment, prompting some further thought. I've written a followup post in which I respond to some of the things they said.